Skip to content

How much VRAM does DeepSeek V4.1 Flash need?

DeepSeek V4.1 Flash needs 431 GB for its weights at 4-bit, 432 GB with an 8K context and 432 GB with 32K.

Parameters
763B
Active per token
16B
Longest context
1M
Released
Sep 2026
License
MIT
Strongest at
Chat, Coding, Agents, Docs

Memory at each precision

Memory DeepSeek V4.1 Flash needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
As released (FP4 + FP8)475 GB476 GB476 GB476 GB
4-bit431 GB432 GB432 GB432 GB
3-bit347 GB348 GB348 GB349 GB
2-bit267 GB267 GB267 GB268 GB
1.6-bit147 GB147 GB148 GB148 GB

One conversation, 16-bit context cache, runtime overhead included.

What runs it

The best way DeepSeek V4.1 Flash runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst reply
RTX 3060 12 GBToo big
RTX 4060 Ti 16 GBToo big
RTX 3090Too big
RTX 4090Too big
RTX 5090Too big
2× RTX 3090Too big
RTX PRO 6000 BlackwellToo big
M4 Max 128 GBToo big
M5 Max 128 GBToo big
Ryzen AI Max+ 395 128 GBToo big
DGX Spark 128 GBToo big
M3 Ultra 256 GB1.6-bit1M82 tok/s9.4 s
M5 Ultra 512 GB3-bit1M62 tok/s3.0 s
H100 SXMToo big
DGX Station GB3001.6-bit1M322 tok/s0.3 s
8× H200 NVLFP4 + FP81M135 tok/s0.1 s

Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.

How to run it

llama.cpp
llama-server -hf unsloth/DeepSeek-V4.1-Flash-1017-GGUF
vLLM
vllm serve deepseek-ai/DeepSeek-V4.1-Flash
MLX (Mac)
mlx_lm.chat --model mlx-community/DeepSeek-V4.1-Flash-MLX-4bit

Questions

How much VRAM does DeepSeek V4.1 Flash need?

DeepSeek V4.1 Flash needs 431 GB for its weights at 4-bit, 432 GB with an 8K context and 432 GB with 32K.

Can DeepSeek V4.1 Flash run on an RTX 4090?

Not on one card. The smallest machine here that runs it is the M3 Ultra 256 GB.

More models